1 /* tac - concatenate and print files in reverse
2 Copyright (C) 1988, 1989, 1990, 1991 Free Software Foundation, Inc.
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 2, or (at your option)
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
18 /* Written by Jay Lepreau (lepreau@cs.utah.edu).
19 GNU enhancements by David MacKenzie (djm@gnu.ai.mit.edu). */
21 /* Copy each FILE, or the standard input if none are given or when a
22 FILE name of "-" is encountered, to the standard output with the
23 order of the records reversed. The records are separated by
24 instances of a string, or a newline if none is given. By default, the
25 separator string is attached to the end of the record that it
29 -b, --before The separator is attached to the beginning
30 of the record that it precedes in the file.
31 -r, --regex The separator is a regular expression.
32 -s, --separator=separator Use SEPARATOR as the record separator.
34 To reverse a file byte by byte, use (in bash, ksh, or sh):
39 #if defined (CONFIG_BROKETS)
40 /* We use <config.h> instead of "config.h" so that a compilation
41 using -I. -I$srcdir will use ./config.h rather than $srcdir/config.h
42 (which it would do because it found this file in $srcdir). */
51 #include <sys/types.h>
62 /* The number of bytes per atomic read. */
63 #define INITIAL_READSIZE 8192
65 /* The number of bytes per atomic write. */
66 #define WRITESIZE 8192
70 static RETSIGTYPE cleanup ();
72 static int tac_file ();
73 static int tac_stdin ();
74 static char *xmalloc ();
75 static char *xrealloc ();
76 static void output ();
77 static void save_stdin ();
78 static void xwrite ();
82 /* The name this program was run with. */
85 /* The string that separates the records of the file. */
86 static char *separator;
88 /* If nonzero, print `separator' along with the record preceding it
89 in the file; otherwise with the record following it. */
90 static int separator_ends_record;
92 /* 0 if `separator' is to be matched as a regular expression;
93 otherwise, the length of `separator', used as a sentinel to
95 static int sentinel_length;
97 /* The length of a match with `separator'. If `sentinel_length' is 0,
98 `match_length' is computed every time a match succeeds;
99 otherwise, it is simply the length of `separator'. */
100 static int match_length;
102 /* The input buffer. */
105 /* The number of bytes to read at once into `buffer'. */
106 static unsigned read_size;
108 /* The size of `buffer'. This is read_size * 2 + sentinel_length + 2.
109 The extra 2 bytes allow `past_end' to have a value beyond the
110 end of `buffer' and `match_start' to run off the front of `buffer'. */
111 static unsigned buffer_size;
113 /* The compiled regular expression representing `separator'. */
114 static struct re_pattern_buffer compiled_separator;
116 /* If non-zero, display usage information and exit. */
117 static int show_help;
119 /* If non-zero, print the version on standard output then exit. */
120 static int show_version;
122 static struct option const longopts[] =
124 {"before", no_argument, &separator_ends_record, 0},
125 {"regex", no_argument, &sentinel_length, 0},
126 {"separator", required_argument, NULL, 's'},
127 {"help", no_argument, &show_help, 1},
128 {"version", no_argument, &show_version, 1},
136 Usage: %s [-br] [-s separator] [--before] [--regex] [--separator=separator]\n\
137 [--help] [--version] [file...]\n",
147 const char *error_message; /* Return value from re_compile_pattern. */
149 int have_read_stdin = 0;
151 program_name = argv[0];
155 separator_ends_record = 1;
157 while ((optc = getopt_long (argc, argv, "brs:", longopts, (int *) 0))
165 separator_ends_record = 0;
173 error (1, 0, "separator cannot be empty");
182 printf ("%s\n", version_string);
189 if (sentinel_length == 0)
191 compiled_separator.allocated = 100;
192 compiled_separator.buffer = (unsigned char *)
193 xmalloc (compiled_separator.allocated);
194 compiled_separator.fastmap = xmalloc (256);
195 compiled_separator.translate = 0;
196 error_message = re_compile_pattern (separator, strlen (separator),
197 &compiled_separator);
199 error (1, 0, "%s", error_message);
202 match_length = sentinel_length = strlen (separator);
204 read_size = INITIAL_READSIZE;
205 /* A precaution that will probably never be needed. */
206 while (sentinel_length * 2 >= read_size)
208 buffer_size = read_size * 2 + sentinel_length + 2;
209 buffer = xmalloc (buffer_size);
212 strcpy (buffer, separator);
213 buffer += sentinel_length;
221 errors = tac_stdin ();
224 for (; optind < argc; ++optind)
226 if (strcmp (argv[optind], "-") == 0)
229 errors |= tac_stdin ();
232 errors |= tac_file (argv[optind]);
235 /* Flush the output buffer. */
236 output ((char *) NULL, (char *) NULL);
238 if (have_read_stdin && close (0) < 0)
239 error (1, errno, "-");
241 error (1, errno, "write error");
245 /* The name of a temporary file containing a copy of pipe input. */
248 /* Print the standard input in reverse, saving it to temporary
249 file `tempfile' first if it is a pipe.
250 Return 0 if ok, 1 if an error occurs. */
255 /* Previous values of signal handlers. */
256 RETSIGTYPE (*sigint) (), (*sighup) (), (*sigpipe) (), (*sigterm) ();
259 #ifdef _POSIX_VERSION
260 struct sigaction oldact, newact;
261 #endif /* _POSIX_VERSION */
263 /* No tempfile is needed for "tac < file".
264 Use fstat instead of checking for errno == ESPIPE because
265 lseek doesn't work on some special files but doesn't return an
267 if (fstat (0, &stats))
269 error (0, errno, "standard input");
272 if (S_ISREG (stats.st_mode))
273 return tac (0, "standard input");
275 #ifdef _POSIX_VERSION
276 newact.sa_handler = cleanup;
277 sigemptyset (&newact.sa_mask);
280 sigaction (SIGINT, NULL, &oldact);
281 sigint = oldact.sa_handler;
282 if (sigint != SIG_IGN)
283 sigaction (SIGINT, &newact, NULL);
285 sigaction (SIGHUP, NULL, &oldact);
286 sighup = oldact.sa_handler;
287 if (sighup != SIG_IGN)
288 sigaction (SIGHUP, &newact, NULL);
290 sigaction (SIGPIPE, NULL, &oldact);
291 sigpipe = oldact.sa_handler;
292 if (sigpipe != SIG_IGN)
293 sigaction (SIGPIPE, &newact, NULL);
295 sigaction (SIGTERM, NULL, &oldact);
296 sigterm = oldact.sa_handler;
297 if (sigterm != SIG_IGN)
298 sigaction (SIGTERM, &newact, NULL);
299 #else /* !_POSIX_VERSION */
300 sigint = signal (SIGINT, SIG_IGN);
301 if (sigint != SIG_IGN)
302 signal (SIGINT, cleanup);
304 sighup = signal (SIGHUP, SIG_IGN);
305 if (sighup != SIG_IGN)
306 signal (SIGHUP, cleanup);
308 sigpipe = signal (SIGPIPE, SIG_IGN);
309 if (sigpipe != SIG_IGN)
310 signal (SIGPIPE, cleanup);
312 sigterm = signal (SIGTERM, SIG_IGN);
313 if (sigterm != SIG_IGN)
314 signal (SIGTERM, cleanup);
315 #endif /* _POSIX_VERSION */
319 errors = tac_file (tempfile);
323 #ifdef _POSIX_VERSION
324 newact.sa_handler = sigint;
325 sigaction (SIGINT, &newact, NULL);
326 newact.sa_handler = sighup;
327 sigaction (SIGHUP, &newact, NULL);
328 newact.sa_handler = sigterm;
329 sigaction (SIGTERM, &newact, NULL);
330 newact.sa_handler = sigpipe;
331 sigaction (SIGPIPE, &newact, NULL);
332 #else /* !_POSIX_VERSION */
333 signal (SIGINT, sigint);
334 signal (SIGHUP, sighup);
335 signal (SIGTERM, sigterm);
336 signal (SIGPIPE, sigpipe);
337 #endif /* _POSIX_VERSION */
342 /* Make a copy of the standard input in `tempfile'. */
347 static char *template = NULL;
348 static char *tempdir;
352 if (template == NULL)
354 tempdir = getenv ("TMPDIR");
357 template = xmalloc (strlen (tempdir) + 11);
359 sprintf (template, "%s/tacXXXXXX", tempdir);
360 tempfile = mktemp (template);
362 fd = creat (tempfile, 0600);
365 error (0, errno, "%s", tempfile);
368 while ((bytes_read = read (0, buffer, read_size)) > 0)
369 if (write (fd, buffer, bytes_read) != bytes_read)
371 error (0, errno, "%s", tempfile);
376 error (0, errno, "%s", tempfile);
379 if (bytes_read == -1)
381 error (0, errno, "read error");
386 /* Print FILE in reverse.
387 Return 0 if ok, 1 if an error occurs. */
398 error (0, errno, "%s", file);
401 errors = tac (fd, file);
404 error (0, errno, "%s", file);
410 /* Print in reverse the file open on descriptor FD for reading FILE.
411 Return 0 if ok, 1 if an error occurs. */
418 /* Pointer to the location in `buffer' where the search for
419 the next separator will begin. */
421 /* Pointer to one past the rightmost character in `buffer' that
422 has not been printed yet. */
424 unsigned saved_record_size; /* Length of the record growing in `buffer'. */
425 off_t file_pos; /* Offset in the file of the next read. */
426 /* Nonzero if `output' has not been called yet for any file.
427 Only used when the separator is attached to the preceding record. */
429 char first_char = *separator; /* Speed optimization, non-regexp. */
430 char *separator1 = separator + 1; /* Speed optimization, non-regexp. */
431 int match_length1 = match_length - 1; /* Speed optimization, non-regexp. */
432 struct re_registers regs;
434 /* Find the size of the input file. */
435 file_pos = lseek (fd, (off_t) 0, SEEK_END);
437 return 0; /* It's an empty file. */
439 /* Arrange for the first read to lop off enough to leave the rest of the
440 file a multiple of `read_size'. Since `read_size' can change, this may
441 not always hold during the program run, but since it usually will, leave
442 it here for i/o efficiency (page/sector boundaries and all that).
443 Note: the efficiency gain has not been verified. */
444 saved_record_size = file_pos % read_size;
445 if (saved_record_size == 0)
446 saved_record_size = read_size;
447 file_pos -= saved_record_size;
448 /* `file_pos' now points to the start of the last (probably partial) block
449 in the input file. */
451 lseek (fd, file_pos, SEEK_SET);
452 if (read (fd, buffer, saved_record_size) != saved_record_size)
454 error (0, 1, "%s", file);
458 match_start = past_end = buffer + saved_record_size;
459 /* For non-regexp search, move past impossible positions for a match. */
461 match_start -= match_length1;
465 /* Search backward from `match_start' - 1 to `buffer' for a match
466 with `separator'; for speed, use strncmp if `separator' contains no
468 If the match succeeds, set `match_start' to point to the start of
469 the match and `match_length' to the length of the match.
470 Otherwise, make `match_start' < `buffer'. */
471 if (sentinel_length == 0)
473 int i = match_start - buffer;
476 ret = re_search (&compiled_separator, buffer, i, i - 1, -i, ®s);
478 match_start = buffer - 1;
481 error (0, 0, "error in regular expression search");
486 match_start = buffer + regs.start[0];
487 match_length = regs.end[0] - regs.start[0];
492 /* `match_length' is constant for non-regexp boundaries. */
493 while (*--match_start != first_char
494 || (match_length1 && strncmp (match_start + 1, separator1,
499 /* Check whether we backed off the front of `buffer' without finding
500 a match for `separator'. */
501 if (match_start < buffer)
505 /* Hit the beginning of the file; print the remaining record. */
506 output (buffer, past_end);
510 saved_record_size = past_end - buffer;
511 if (saved_record_size > read_size)
513 /* `buffer_size' is about twice `read_size', so since
514 we want to read in another `read_size' bytes before
515 the data already in `buffer', we need to increase
518 int offset = sentinel_length ? sentinel_length : 1;
521 buffer_size = read_size * 2 + sentinel_length + 2;
522 newbuffer = xrealloc (buffer - offset, buffer_size) + offset;
523 /* Adjust the pointers for the new buffer location. */
524 match_start += newbuffer - buffer;
525 past_end += newbuffer - buffer;
529 /* Back up to the start of the next bufferfull of the file. */
530 if (file_pos >= read_size)
531 file_pos -= read_size;
534 read_size = file_pos;
537 lseek (fd, file_pos, SEEK_SET);
539 /* Shift the pending record data right to make room for the new. */
540 bcopy (buffer, buffer + read_size, saved_record_size);
541 past_end = buffer + read_size + saved_record_size;
542 /* For non-regexp searches, avoid unneccessary scanning. */
544 match_start = buffer + read_size;
546 match_start = past_end;
548 if (read (fd, buffer, read_size) != read_size)
550 error (0, errno, "%s", file);
556 /* Found a match of `separator'. */
557 if (separator_ends_record)
559 char *match_end = match_start + match_length;
561 /* If this match of `separator' isn't at the end of the
562 file, print the record. */
563 if (first_time == 0 || match_end != past_end)
564 output (match_end, past_end);
565 past_end = match_end;
570 output (match_start, past_end);
571 past_end = match_start;
573 match_start -= match_length - 1;
578 /* Print the characters from START to PAST_END - 1.
579 If START is NULL, just flush the buffer. */
582 output (start, past_end)
586 static char buffer[WRITESIZE];
587 static int bytes_in_buffer = 0;
588 int bytes_to_add = past_end - start;
589 int bytes_available = WRITESIZE - bytes_in_buffer;
593 xwrite (1, buffer, bytes_in_buffer);
598 /* Write out as many full buffers as possible. */
599 while (bytes_to_add >= bytes_available)
601 bcopy (start, buffer + bytes_in_buffer, bytes_available);
602 bytes_to_add -= bytes_available;
603 start += bytes_available;
604 xwrite (1, buffer, WRITESIZE);
606 bytes_available = WRITESIZE;
609 bcopy (start, buffer + bytes_in_buffer, bytes_to_add);
610 bytes_in_buffer += bytes_to_add;
621 xwrite (desc, buffer, size)
626 if (write (desc, buffer, size) != size)
628 error (0, errno, "write error");
633 /* Allocate N bytes of memory dynamically, with error checking. */
644 error (0, 0, "virtual memory exhausted");
650 /* Change the size of memory area P to N bytes, with error checking. */
660 error (0, 0, "virtual memory exhausted");